UNIVERSITY OF BUCHAREST DEPARTAMENT OF SYSTEMS ECOLOGY AND SUSTAINABILITY Tel AQUAMONEY CASE STUDY REPORT Islands of Braila complex (Inner Danube Delta) Prof. Dr. Angheluta VADINEANU, Dr. Nicoleta GEAMANA, PhD Student Teodora PALARIE 25 November 2008 Table of Content 1. Introduction 2. Description of the case study 2.1. Location of the case study area 2.2 Water system characteristics 2.3. Short characterization of water use and water users 2.4. Main water management and policy issues in the context of the WFD 3. Set up of the survey 3.1. Questionnaire design (common) 3.2. Sampling procedure and response rate 4. Valuation results 4.1. Respondent characteristics and sample representativeness 4.1.1. Demographic characteristics 4.1.2. Socio-economic characteristics 4.1.3. Water use characteristics 4.2. Public perception of water management problems 4.3. Estimated economic values for water resource management 4.4. Factors explaining economic values for water resource management 4.5. Total Economic Value 5. Conclusions 6. Best practice recommendations 2 1. Introduction As part of the AquaMoney deliverables, the current report addresses the following project objectives: test the practice-oriented guidelines for assessing environmental and resource costs and benefits (ERCB) developed in the first year of project activity, by carrying out pilot case studies in 10 different European river basins; analyze experiences in the pilot case studies and translate these into practical policy guidelines. The main objective of this report is to assess public perception of the values associated with river restoration projects in terms of flood control and water quality improvements in the Romanian LTSER site Braila Islands, part of the international Danube Basin. By using choice experiment and contingent valuation, the economic values of river restoration are elicitated in order to estimate non-market values which can be used, together with potential market benefits such as the avoided costs of flood damage or water purification, to justify investments in Danube river restoration projects to achieve the environmental and ecological objectives of the WFD based on economic welfare considerations. Danube, the second largest river in Europe, has been strongly modified during the last 5-6 centuries by different canalization and embankments, navigation and hydropower works. This is also the case of Braila Islands Danube sector where the Romanian study took place. Therefore, with the WFD in placed the objective of achieving a “good ecological potential” for heavily modified water bodies, according with Article 4, requires water body specific measures including ecological restoration. The whole arsenal of measures that might insure achieving this ecological objective is governed by the costs of such improvements and the scope for time derogations and the setting of less restrictive targets because of disproportionate costs. At the same time, multiple benefits will reside from WFD implementation like improving the ecological status of water bodies and wetlands including the pollution reduction, decreasing the frequency and extent of floods and droughts, encouraging participation in water-based recreation, developing recreational activities, biodiversity conservation etc, from which only some will be quantifiable in monetary terms. In this context, there is an urgent policy requirement for better understanding of nonmarket benefits and costs of WFD implementation, in order to scientifically underpin WFD implementation strategies. 3 2. Description of the case study 2.1. Location of the case study area Braila Islands is a large Long Term Socio-Ecological Research site, situated in the South-East of Romania, that extends over 2600km2 and corresponds to a 78km long Danube sector that stretches between Harsova (kilometer 253) and Braila (kilometer 175) cities. This socio-ecological system is inhabited by near 300,000 people and comprises heavily modified ecosystems (e.g. Big Island of Braila) but also systems under a natural functional regime (e.g. Small Islands of Braila), being of a crucial natural and socio-economical value. Braila Islands Location of the case study area 27 48’36" E / 28 15 ’52" E 45 25’26” N / 44 35’ 18” N Fig.1. Location of the Braila Islands in the Lower Danube Basin. 2.2 Water system characteristics The Danube river in the Braila Islands section has been ranked as a heavily modified water body according to criteria 2.1 (embankment works) due to the hydro-technical works on 79% of the river stretch sector and a candidate to “heavily modified” according with the WFD criteria 2.2 (regulation works) as a result of dredging of 21% of the river bed for intensive navigation. The main remnant of the natural floodplains consists in the wetlands from the Small Island of Braila 4 Natural Park (SIBr) with a total surface of 210 square kilometers and the floodplains between the riverbanks and dikes of almost 93 square kilometers. 2.3. Short characterization of water use and water users 2.3.1. The main uses related to water in the region are: Agriculture – 73.46% of the total area is represented by agricultural land. In 2004, of the total 191,000 ha representing the agricultural surface, 71,960 ha (37.68% from the total agricultural area in Braila Islands) was irrigated according with water users associations data1. At that time, the widest irrigated area was represented by the Big Island of Braila with 64,663 ha (97% of the area) due to its predominant agricultural character. Navigation – The Danube segment between Harsova and Braila (Braila Islands area) is part of an important sector (Calarasi-Braila) of the Pan-European corridor no. VII. In the case-study section freight and passengers transport is done through Braila Harbour. Because, Braila port assures the connection between the fluvial and maritime Danube, in 2006 the cargo traffic consisted in approximatively 940,000 tones from which the maritime traffic accounted for about 220,000 tones and 720,000 tones for fluvial traffic according with the Union of Romanian Inland Ports data2. As draught is severely affecting navigation on the fluvial Danube (e.g. the draught from 2003), an ISPA project was designed to assure stabilization of the fairway through riverbed rehabilitation and improving of hydromorphological conditions in some critical points of the Calarasi-Braila sector. Therefore after 2012 it is expected that the cargo transport to strongly increase. Recreation and tourism – there are 3 touristical attraction areas in the Islands of Braila: Small Island of Braila Natural Park, Macin Natural Park and Lacu Sarat Spa resort. Nevertheless local people use more leisure areas from the region from the Braila beach (a recreation area that was created in the Big Island of Braila, on the right arm of the Danube to small recreational areas without facilities) Industry and households – there are two water plants in the area, one in Braila and one in Chiscani, which are supplying the end-users with drinking water extracted from Danube. 1 Ministry of Agriculture and Rural Development web-page, Journal of Irrigation Water Users Association registered in the National Register of Water Users Association, December 2004, downloaded from: http://www.maap.ro/pages/page.php?self=01&sub=0106&art=0607&var=010603 2 Union of Romanian Inland Ports web-page, Traffic Data 2000-2006, downloaded from: http://www.danubeports.ro/trafic_braila.html 5 2.3.2. Water users The main water users are grouped in the three economic sectors: primary (agriculture, fishing, forestry and minig), secondary (manufactoring and construction) and tertiary (services). The only form of organization of the water users takes place in agriculture where Irrigation Water Users Associations (Asociatii ale Utilizatorilor de Apa pentru Irigatii) have been establisehed. The primary sector is relatively low represented, despite the fact that the area is primarily an agricultural one, due to the fact that the majority of farms is represented by small subsistance farms, which do not account for companies turnover and that some of the big farms in the area are registered in other counties. Secondary sector 39% Tertiary sector 54% Commerce 44% Primary sector Secondary sector Transport and tourism Commerce Other services Primary sector 7% Transport and Other services 5% tourism 5% Fig.2: Economic sectors Islands of Braila according with the registered companies’ turnover3 2.4. Main water management and policy issues in the context of the WFD The need to adapt the water strategy and management to the trend of increasing frequency and intensity of droughts and floods; Wetlands restoration, up to one half of the current agricultural polders existing in the physical structure of the Braila Islands; Agricultural landscape planning for multifunctional farming system which may allow for effective diffuse pollution control, habitat connectivity and biodiversity conservation; Building and improving the water drainage system within the remaining agricultural polders; Rehabilitation of the flood defence system; Dredging river bad and improving navigation; Improving water flow inside SiBr; Development of the water supply infrastructure for about 22 per cent of population, living in the rural area around BrI; Efficient and effective waste water treatment infrastructure development. 3 Statistical data provided by the National Statistics Institute 6 3. Set up of the survey 3.1. Questionnaire design The questionnaire was developed after several meetings, discussions and pre-tests and consisted of four main parts: Perceptions and attitudes. The first part of the questionnaire contained questions about respondents’ general perceptions and environmental attitudes. Respondents were asked, for example, about types and frequency of recreational activities in the catchment area and how often they visit the case study area. This section also captured people’s perceptions about water quality and water quality evolution over the last ten years. Choice Experiment. In the second part, respondents were asked to state their choices using four different choice sets. In the introduction to the choice experiment a map of the location of the river restoration area was show to each respondent. The maps were based on CORINE LANDCOVER 2000 (shape file 1:100000). The major types of ecosystems were derived from Corine classes’ level 3 and provided information about human settlements, agricultural systems, forests and meadows, wetlands and freshwater ecosystems. The CE was followed up with a debriefing question and respondents who opted out (i.e. chose not to select one of the alternatives) four times were asked why they chose as they did. Contingent Valuation. The CE was followed up by a CV-question on ecological restoration. Participants were asked to state their maximum willingness to pay in order to help finance (largely unspecified) restoration measures which they were told would change the ecological status and/or recreational potential of the area. Demographic/socio-economic data. The final part of the questionnaire was focused on gathering data on respondents’ demographic and socio-economic status (income, age, number of children, current work status, education, etc.). 3.1.1. Design and Implementation of the choice experiment In order to estimate and justify expenses for river restoration programs ecologists consider to be beneficial (in order to assist the decision-making processes), in the present study a choice experiment (CE) was chosen to value ecological restoration and to estimate the WTP for certain restoration management programmes. The design consisted of two exclusive categories of benefits: the impact of river restoration on floodwater storage and the corresponding reduction of flood risk, and the river’s nutrient retention capacity and hence water quality. Therefore, the CE was composed of three attributes (flood frequency, water quality and cost of the option) and respondents were asked to choose between the current situation and two alternatives. Respondents were told in the introduction that river restoration measures can positively affect water quality and flood frequency. The degree of restoration of the river (towards a more natural state) is connected to the degree of water quality improvement and flood frequency decrease that can be expected. Water quality was described in terms of variety of aquatic life and recreational uses such as swimming, booting and fishing. A selection of multi-colored pictograms was used to assist respondents to visualize different quality levels, starting from moderate to good and very good water quality (Fig.3). The differences between the levels were explained in detail. 7 Option A Option B Status Quo Once every 25 Once every 25 Once every 5 years years years Flood frequency Good Moderate Very good Water quality Increase water bill in €3 (25 Cent month) I choose: (Please tick as Option A appropriate) € 10 / (83 Cent month) Option B / No additional payment Neither Fig.3. Example choice card Flood frequency was defined as the probability to cause damage (financial losses) to communities, agricultural and industrial uses in the areas downstream of the river restoration and re-naturalization measures, with the four levels: 5, 25, 50 and 100 years. The lowest level for both attributes, water quality and flood frequency corresponded to the status quo. The monetary attribute payment vehicle was specified as an increase in the respondents’ water bill to fund the water management programme (in the form of an annual contribution on top of the water bill). The payment levels used in the choice experiment were equivalent amounts of 3, 10, 30 and 50 € expressed in Romanian Lei. In order to combine the levels of the attributes into a number of options a fractional factorial design was used. 32 choice sets were assigned to 8 blocks such that each respondent was confronted with a randomly selected four choice set. 3.1.2. Design of the contingent valuation scenarios In the study, the contingent valuation method consisted of asking respondents about their willingness to pay for increasing the size of natural areas along the river - from the actual situation to an ecologically enhanced situation. Respondents were told that, with restoration measures, wetlands and forests could be connected to the Danube river which would lead to a more natural landscape with water flowing not only through the main channel but also through adjacent creeks and ponds (Box 1). Respondents were told that currently about 20 % (210 km2) of the former wetlands are still in a natural shape. It was also mentioned that the reference state of the area (1056 km2) contained a large number of shallow lakes, ponds and marshes, linked to 8 each other by natural or man made channels and the entire network of freshwater /wetland ecosystems was connected to the Danube river arms. Box 1: Introduction of the CV-question As described before, the Danube River is heavily modified in many places. Today approximately a quarter of the river is still connected the surrounding floodplains and wetlands and the river banks are still in a natural state (SHOW MAP OF THE CURRENT SITUATION). Restoration measures would connect the river again to the floodplains and the wetlands as they were originally before the changes made to the river and river banks. As a result of river and floodplain restoration the landscape will look more natural, with water flowing also through adjacent creeks and ponds. This more natural state has a positive effect on nature and the variety of plant and animal species found in the catchment. Plans exist to restore half (50 percent) (alternatively 90%) of the former wetlands in the Braila Islands catchment back into their original natural state as shown on the map (SHOW MAP), and connect the river again with the floodplains and wetlands. The respondents were explicitly told that for each scenario they should state the maximum amount they would be willing to pay on top of their annual water bill in order to restore a certain degree of the river bank. We used an open-ended format (a payment card) to elicit individuals’ maximum willingness. The payment card showed 29 values ranging from €0 to €250. Additionally, the payment card offered the options “more than € 250, namely …”, “other amount, namely…” and “I don’t know”. The WTP question was formulated as follows: “Can you tell me with the help of this card how much you are willing to pay MAXIMUM on top of your yearly water bill over the next 5 years for the restoration of half (alternatively 90 %) of the modified river banks in the Braila Islands catchment area back into their original natural state as shown on the map?” Those respondents who were not willing to make a financial contribution to restoration measures were asked to state why. In addition, these respondents were confronted with a series of statements (e.g. “It is the task and responsibility of the government to protect the rivers” or “The environment has the right to be protected irrespective of the costs of the society.”) to identify and categorize protest bidders. 9 3.2. Sampling procedure and response rate The main survey was carried out between 12th and 17th of November 2007 following a random sampling procedure (every 10th person in the urban area – city of Braila, every 5th person in the rural area - 19 settlements situated on the right and left arm of Danube) and the sample size has included 851 asked persons, from which only 61% (519 persons) accepted to complete the questionnaire and 39% (332 persons) have refused to answer the questionnaire. In the urban area almost 44% (316 persons) of the contacted person refused to participate in the survey. A frequent motivation was: “I’ve already filled in other questionnaire” which allow us to assume that this was a simple excuse for saying No or that they had been already involved in other investigations based on questionnaires. The last assumption seems to be more reliable because frequent investigations dealing with the assessment of the credibility of politicians/ parties (preparation work of the election campaign for European Parliament, local authorities and National Parliament) were carried out at that moment. In rural areas the majority of refuses (14,5% of the contacted persons) came from old people (over 65 years old) who justified their attitude by lack of trust that their opinion will be taken into account by decision-makers. The respondents sample consisted in 49,13% female and 50,87% male and, 78,8% of the respondents (409 people) living in the urban area and 21,2% (110 respondents) in the rural area. The interviews’ locations were public places with high pedestrian traffic: main squares, parks, in front of shopping centers and shops, Mayoralties, Postal offices. A special attention was given to the composition of the sub-sample in order to mimic the local population structure including sex ratio (e.g. 1/1), age classes, level of education, income categories. 4. Valuation results 4.1. Respondent characteristics and sample representativeness 4.1.1. Demographic characteristic a) Gender The percentage of males wass 51% in our sample, while in the region it is 48.3%. According to the t-test it is representative (see below). T-TEST /TESTVAL=0.483136 /MISSING=ANALYSIS /VARIABLES=Sex /CRITERIA=CI(.9500). One-Sample Statistics Sex N Mean Std. Deviation Std. Error Mean 519 ,51 ,500 ,022 10 One-Sample Test Test Value = 0.483136 95% Confidence Interval of the Difference Sex t df Sig. (2-tailed) Mean Difference Lower Upper 1,162 518 ,246 ,026 ,07 -,02 Since Sig. is larger than 0.246 >0.05, and the Confidence Interval of the Difference includes the zero value, we come to the conclusions that there is no significant difference in the variance between the sample and the local population, therefore the sample is representative from the gender perspective. b) Age The age structure of the sample is quite heterogeneous and covers all categories of age classes, with a better representation for the 28-57 years interval (see Figure 3). AGE 120 108 117 108 100 80 Number sample 79 57 60 Series1 39 40 20 11 0 18-27 28-37 38-47 48-57 58-67 68-77 78-87 Fig.4. Age classes of the sample With a mean difference of -1.06 and a Sig. value of 0.121, the t test shows that the average age of the sample (44.47 years) is representative for the adult local population (45.53). As the survey addressed only people of 18 and over, the average age for the adult local population (45.53) is bigger than the average age for the local population that takes into account all class ages (39.54) 11 T-TEST /TESTVAL=45.53 /MISSING=ANALYSIS /VARIABLES=Age /CRITERIA=CI(.9500). One-Sample Statistics N Age Mean 519 Std. Deviation 44,47 Std. Error Mean 15,585 ,684 One-Sample Test Test Value = 45.53 95% Confidence Interval of the Difference t Age df -1,552 Sig. (2-tailed) Mean Difference 518 ,121 Lower -1,062 Upper -2,41 ,28 c) Household size In the region the average household size is 3.11, while in our sample it is 3.1. According to the ttest it is representative (see below). T-TEST /TESTVAL=3.11 /MISSING=ANALYSIS /VARIABLES=nopers /CRITERIA=CI(.9500). One-Sample Statistics nopers N Mean Std. Deviation Std. Error Mean 519 3,10 1,237 ,054 One-Sample Test Test Value = 3.11 95% Confidence Interval of the Difference nopers t df Sig. (2-tailed) Mean Difference Lower Upper -,216 518 ,829 -,012 ,09 -,12 12 With a Sig. value of 0.829 the sample is representative for the local population. d) Urban-rural ratio The population in the region is mainly located in urban areas (Braila, Harsova and Macin), therefore 80% of the people are living in the city, while in our sample is 78.8% of the population. The t-test shows that the sample is representative: T-TEST /TESTVAL=0.8 /MISSING=ANALYSIS /VARIABLES=urban /CRITERIA=CI(.9500). One-Sample Statistics urban N Mean Std. Deviation Std. Error Mean 519 ,79 ,409 ,018 One-Sample Test Test Value = 0.8 95% Confidence Interval of the Difference urban t df Sig. (2-tailed) Mean Difference Lower Upper -,665 518 ,506 -,012 ,02 -,05 with a Sig.value of 0.506 and a confidence interval containing the zero value. 13 4.1.2. Socio-economic characteristics Due to the limited statistical data available in the national and regional statistics regarding socio-economic characteristics, an extensive examination of the representativeness of the sample is impossible. For the occupation structure the only available data in the national statistics are the numbers of unemployed people at the county level. Having in mind the fact that more than 80% of the case-study population lives in the Braila County we used Braila data for the sample comparison, keeping in mind though the fact that by doing this we introduce a new level of uncertainty in the analysis. In the Braila County there are 8733 unemployed people, representing 2.38% of the county population, which is similar to the 3% unemployed people from the sample. Still this information is not significant to come to the conclusion that the sample is or not representative for the local population. The sample occupation structure is as following: Occupation PFA 1%2% 9% Employed 22% Part-time employed Student Unemployed House keeper 5% 3% 5% Pensioner 52% 1% Handicap Other Fig. 5. Occupation structure of the sample The education structure of the sample shows that a great number of the respondents (almost 40%) have a minimal education (primary school and vocational) which can be explained by the fact that the area is primarily an agricultural area, and also the big percentage of people with technical background which is due to the industry clustering in the city of Braila. Frequency Valid primary school 92 professional 104 high school 181 college 106 (technical) university 33 Other 3 Total 519 Percent 17.7 20.0 34.9 Valid Percent 17.7 20.0 34.9 Cumulativ e Percent 17.7 37.8 72.6 20.4 20.4 93.1 6.4 .6 100.0 6.4 .6 100.0 99.4 100.0 14 Table 1. Education structure of the sample Education 6% 1% 18% School 20% Vocational school High school College 20% University Other 35% Fig.6. Education structure of the sample 4.1.3. Recreational water use characteristics Despite the fact that the Small Islands Income of Braila is the second most valuable wetland area in Romania, after the Danube Delta (the coastal delta), nevertheless there are about 55% of the respondents who claim they have never visited the park. In this category are also included those 0.58% who have been in the area but not for recreational purposes. From the respondents0-250/month who have 251-500 visited the area, the great majority are going there at least once a year. There are only 4% of the 501-750 respondents who are visiting4.24% the Natural 1.54% Park on weekly bases. 750-1000 0.19% Number of visits to the Small Islands of Braila Natural Park 1001-1500 1501-2000 2001-2500 2501-3000 13.29% 30% 31.41% 11.75% 55% 8% 3% 4% 36.99% I never visit the Small Island of Braila I visit it at least once a week I visit it at least once a month I visit it at least 4 times a year I visit it at least once a year Fig.7. Income structure of the sample 15 The average net income of the respondents is about 5640 euros/household/year, but there are big differences between the households per year income in the rural areas compared to the one in the urban area. 180 160 140 120 100 Rural 80 Urban 60 40 20 0 1- 250 251-500 501-750 751-1000 10011500 15012000 20012500 25013000 Fig.8. Household net income (euros/month) in urban and rural areas in Braila Islands No respondent from the rural area reported a household net income per month larger than 1500 euros, while about 54.5% of the rural respondents fit into the first income category (1-250 euros/month). In both rural and urban areas about one third of the respondents had a household income between 251 and 500eurs /month, but in the urban areas more than one third (37.4%) of the respondents had an income greater than 501 euros/household/year compared to 10% in the rural areas. 4.1.3. Water use characteristics The most common recreational activities undertaken by respondents are walking and relaxation, with almost 50% of the respondents actually doing these activities quite often. Recreational activities undertaken by respondents 100% 90% 80% 70% 60% never 50% sometimes 40% often 30% 20% 10% 0% fishing swim boat walks othsport relaxing wildobs picknik caffe dogwalk kidsplay Fig.9. Recreational activities undertaken by respondents 16 The most unpopular activity seems to be dog walking with less than 10% of the respondents who are dog walking in the vicinity of Danube as a recreational activity. This is explainable by the fact that in rural and also in urban areas that are still developing like Braila dogs are not seen as pets but as farm animals, just like chickens and pigs, which have a specific role of guarding the farm. Swimming and fishing seem to be similarly popular among respondents (with 40% of the respondents doing the activities sometimes or often), which can be explained by the fact the majority of respondents live in urban areas where there not proper places arranged for these activities. 4.2. Public perception of water management problems 4.2.1. Flood experiences Floods and their control are considered an important problem by the respondents regardless the fact that only 8% of the respondents had experienced floods problem during their life time, while the other 92% had never suffered any losses caused by floods. From the people who had experienced floods the majority had suffered due to agricultural land flooding. Importance of flood control 2% 3% 19% not important at all not important somewhat important 76% very important Fig.10. Importance of flood control Comparing the importance attributed by the respondents to water quality to the one ascribed to flood control, it seems that people view floods and water quality similarly important. Even though, there are more people who do not consider important water quality compared to flood. Still, the majority (72%) consider water quality as being very important (see Fig. 11). 17 Importance of the water quality 5% 23% Not important at all Somewhat Very important 72% Fig.11. Importance of the water quality People perception about water quality in the Danube differs from reality, as a large fraction of the sample (45%) considers water quality poor. Actually water quality is moderate to good, with some pollution hot-spots, the main problem in the area being the large hydromorphological modifications (canalization, dredging, and embankments) of the river arms that affected the natural functioning of the system. Almost 45% of the sample views water quality closer to reality as being moderate and good, while only 1% believes the water is very good. Water quality in Danube 1% 9% Poor 15% 45% Moderate Good Very good 30% Don't know Fig.12. Water quality in Danube Evolution of water quality seems to have deteriorated in the last 10 years in the public perception, as 62% of the respondents have indicated it, contrary to realty that shows an improvement of water quality in the last years. 18 Evolution of the water quality 6% 11% 21% Improved No change Deteriorated Don't know 62% Fig.13. Evolution of the water quality in the last 10 years 4.3. Estimated economic values for water resource management 4.3.1. Public willingness to pay for ecological restoration 4.3.1.1. Results from the contingent valuation In order to analyze and compare people’s willingness to pay for restoration projects, people were asked for both WTP for the two different scenarios: 50% and 90% restoration of the former wetlands (see Fig. 14 and 15). Financial contribution for ecological restoration projects (50% ) Series1 21 0 0 0 0 20 an 0 20 0 0 m or e th 10 0 50 75 3 40 7 30 15 20 10 5 3 1 0 180 169 160 140 121 104 120 100 80 51 60 30 40 13 20 0 Fig. 14. Financial Contributions in Euros/year for ecological restoration projects (50%) 19 Financial contribution for ecological restoration projects (90% ) Series1 5 1 0 0 0 0 20 0 22 75 7 40 20 10 3 0 200 190 180 160 140 120 90 97 100 80 51 60 32 24 40 20 0 Fig.15. Financial Contributions in Euros/year for ecological restoration projects (90%) One third of the respondents were zero bidders for the 50% restoration scenario, and the number increase by 4% in case of the 90% scenario. The difference consisted in people who believed that the 90% scenario is not possible or is not desirable, as it will affect the agricultural activities in the area, by transforming large polders into wetlands. According with the CV analysis results, local population is willing to pay for restoration projects, despite the low level of income (mean income = 5640Euro/year/household) ~ 1% of the mean annual income of the sample. The difference in bids for the 50 % and 90 % restoration projects in the CV is not significant (the mean of the bids is: for 50 % - 46.97 Euro/year and for 90% restoration - 52.93 Euro/year). Taking into account that in the area are about 103.460 households the total economic value associated with the two restoration projects are: 50% Braila Islands 4.86 90% 5.48 Table 1: Estimated total economic value (TEV) in million Euros per year for 50% and 90% restoration scenarios in Braila Islands: 20 4.3.1.2. Results from Choice Experiment +---------------------------------------------+ | Discrete choice (multinomial logit) model | | Maximum Likelihood Estimates | | Dependent variable Choice | | Weighting variable ONE | | Number of observations 2076 | | Iterations completed 4 | | Log likelihood function -2254.513 | | Log-L for Choice model = -2254.5131 | | R2=1-LogL/LogL* Log-L fncn R-sqrd RsqAdj | | No coefficients -2280.7191 .01149 .01054 | | Constants only. Must be computed directly. | | Use NLOGIT ;...; RHS=ONE $ | | Response data are given as ind. choice. | | Number of obs.= 2076, skipped 0 bad obs. | +---------------------------------------------+ +---------+--------------+----------------+--------+---------+---------+ |Variable | Coefficient | Standard Error |b/St.Er.|P[|Z|>z] | Mean of X| +---------+--------------+----------------+--------+---------+---------+ ASC -.3048910310 .10698173 -2.850 .0044 PR -.7236477008E-02 .21170583E-02 -3.418 .0006 FL .5364740473 .46088788 1.164 .2444 QUL .2461949719 .44019830E-01 5.593 .0000 Table 2. Discrete choice (multinomial logit) model Maximum Likelihood Estimates A multinomial logit model - using a maximum likelihood (ML) estimation method - was employed to analyze the choice data. The ML method calls for assumptions about probability distribution functions, such as the logistic function and the complementary log-log function. All coefficients are correctly signed according to a priori expectations, meaning the price is negative and the water quality and flood frequency positive. The probability for the water quality attribute to have a coefficient value (log of odds) of 0.25 is 100% according to the model, but the probability for the flood frequency coefficient to be 0.54 is unfortunately only 76% which is far under the confidence level of 95% (using an alpha value of 0.05), therefore the flood coefficient is not statistically significant. Implicit prices for river restoration attributes can be derived by comparing the ratio between the coefficients for each attribute and the coefficient for the monetary attribute, everything else being equal. On average, respondents are willing to pay for the flood frequency improvements about 74.135 euros/year and for water quality improvements 34.021 euros/year. Still correlating the p-values with the implicit prices we can actually come to the conclusion that while it is almost certain that local population of Braila Islands will pay 34.021euros/year for water quality improvements, we can not be confident that the choice of paying 74.135euros/year for flood frequency reduction is not purely accidental and that is actually reflecting their WTP for this attribute. 21 +---------------------------------------------+ | Discrete choice (multinomial logit) model | | Maximum Likelihood Estimates | | Dependent variable Choice | | Weighting variable ONE | | Number of observations 2076 | | Iterations completed 4 | | Log likelihood function -2253.812 | | Log-L for Choice model = -2253.8119 | | R2=1-LogL/LogL* Log-L fncn R-sqrd RsqAdj | | No coefficients -2280.7191 .01180 .01013 | | Constants only. Must be computed directly. | | Use NLOGIT ;...; RHS=ONE $ | | Response data are given as ind. choice. | | Number of obs.= 2076, skipped 0 bad obs. | +---------------------------------------------+ +---------+--------------+----------------+--------+---------+---------+ |Variable | Coefficient | Standard Error |b/St.Er.|P[|Z|>z] | Mean of X| +---------+--------------+----------------+--------+---------+---------+ ASC .6615037582E-02 .97892972E-01 .068 .9461 PR -.5721932418E-02 .28482023E-02 -2.009 .0445 FLOW100 -.1601268706 .12510458 -1.280 .2006 FMED50 -.1284072460 .97298083E-01 -1.320 .1869 FHIGH25 -.3960500109E-01 .90066659E-01 -.440 .6601 QGOOD .3089792650 .95069101E-01 3.250 .0012 QVERGOD .4989618767 .91652918E-01 5.444 .0000 Table 3: Discrete choice (multinomial logit) model Maximum Likelihood Estimates Marginal change in: Braila Islands Flood frequency 0 (not significant) Water quality conditions moderate good 50.7 (23.6) moderate very good 81.2 (31.1) Note: standard errors between brackets. Table 4: Attribute implicit prices (€/household/year) The values presented in Table 4 represent marginal WTP: what a household is willing to pay, for a reduction of the flood return period with one year and for a change in water quality from moderate to good and very good conditions. In the case of floods, the WTP seems to be zero, while for the water quality improvements the sums are extremely large, but holding huge standard errors. 22 4.4. Factors explaining economic values for water resource management Variable Mean fixed effects ASC Flood frequency Estimate 0.325 -0.004 s.e. 0.119 0.002 p< 0.006 0.029 Water Q Good Water Q Very Good Water QG x Perception Water QVG x Perception Water QG x Future visit Water QVG x Future visit Mean random effects 0.174 0.455 -0.003 -0.007 0.463 0.447 0.151 0.134 0.003 0.002 0.168 0.136 0.248 0.001 0.233 0.001 0.006 0.001 Cost price Cost price x income Standard deviation Cost price Cost price x income Model fit Log Likelihood Adjusted R square N -0.042 0.014 0.007 0.002 0.001 0.001 0.016 0.006 0.013 0.003 0.198 0.030 -1805.617 0.056 1996 Table 4: Estimated mixed logit choice models According with the very low R square values, the model does not succeed to explain the choice of one of the two restoration alternatives proposed by the study. Still, what seems to have an influence on people’s choices are the higher expectations of people related to the water quality level, and the intention of future visits in the area. Even though the flood frequency estimate is negative, contrary to what it was expected, its value is very close to zero, which means that it has no influence on people’s choices. Still the negative sign is a surprise that can be explained by the fact that despite the effort put into the wording of the survey, still it was not clear enough for the respondents the difference between the floods as an attribute, which was described as a negative event causing economic damage, and the natural floods that will reside from a natural configuration and functionality of a river systems. Probably by perceiving floods as a beneficial natural process that creates habitats for fish and birds spawning, favors the regeneration of fertile soils for agriculture and other resources, etc. respondents felt that a decline in flood frequency to be a negative result that will diminished all the benefits generated by flooding and therefore as a decrease in welfare. Another possible explanation can reside from people’s lack of understanding on how ecological restoration and therefore flooding of some areas can solve the catastrophic flood problems. Maybe asking people for their willingness to pay for restoration projects sounded more like asking someone to “fight fire with fire”, or in our case to “fight floods with floodings”. 23 4.5. Total Economic Value Policy scenario Braila Islands Flooding Water quality WTP s.e. 1 Once every 25 yrs Good 9.31 2.26 2 Once every 50 yrs Good 7.52 2.63 3 Once every 25 yrs Very good 22.56 4.28 4 Once every 50 yrs Very good 20.77 4.49 5 Once every 100 yrs Very good 17.19 5.59 Table 5: Consumer surplus welfare measures (€/household/year) for different policy scenarios From the above table the WTP for very good quality is higher than for good quality, showing people interest and concern about water quality. For the first two scenarios characterized by good water quality and a flood frequency decrease from 25 years (1st Scenario) to 50 years (2nd scenario) the difference in WTP is not significant. Also a decreasing WTP is observed whenever the flood frequency reduces. This can be explained by the fact that people feel less concerned and are less willing to take responsibility of paying for long term results. The large differences between WTP for scenario 1 and 3, and also for scenarios 2 and 4 show that the most stringent for the people is water quality, which has an impact on people’s day-to-day life. At the same time, people are not willing to pay for “probable” events like floods, especially when there are on larger time scales (almost generation periods). Good Very good Whole country 308.5 462.7 Distance correction 91.6 176.4 Distance & income correction 112.8 155.6 Market size (km)* 59 86 * Distance where value reduces to €0. Table 6: Estimated total economic value (TEV) in million Euros per year for good and very good water quality in Romania based on different aggregation procedures From the above table we can observe that in Romania the TEV for good water quality reduces by 70 percent when accounting for distance-decay. Secondly, accounting for distance-decay and income variation within the boundaries of the market size the TEV increases due to the positive income effects of a number of large cities near the Danube river (including Bucharest), which more than compensate for the negative impact of distance-decay. However, extending the market size for very good water quality (from 59 to 86 km) and accounting for both distance-decay and income effects results in a similar reduction of the TEV of around 10 percent. 24 5. Conclusions The study carry in the Braila Islands LTSER site was the first exercise of economic evaluation in the field of water management using the methodology of CE in Romania. Still, the little experience held by the Romanian team was compensated by the close coordination and support of the project leaders. The Romanian team exercise proves once more that economic evaluation of ERCB is quite a difficult task which requires finances, time and expertise. These are probably the most important limitations of the valuation process, showing the need of developing reliable studies and methodologies for benefit transfer across river basins. The study results show that there is a great concern from the local people related to water quality and that there is no willingness to pay for long term results (decrease in flood frequency). By comparison with the other two Danube basin countries involved in the AquaMoney project (Austria and Hungary), this situation is not specific to all danubian countries which raises again the issue of transferability and its limitations. As people perception related to water quality differs from reality, as people believe that the quality is worse than it really is, show that there is a great need to inform and communicate with the general public. Communication can also help decision-makers prioritize the measure that need to be taken and which are the major aspects that need to be emphasized in the public consultation process (e.g. water quality rather than floods). 25 6. Best practice recommendations As monetary valuation methods can provide relevant input for decisions related to Article 9 of the WFD and is most likely to play a decisive role is the decision on exemptions on the grounds of disproportionate costs, the need of developing and explaining the methodologies for the economic valuation is crucial. Therefore in order to avoid any pitfalls the experience gained in the Aquamoney project might prove to be extremely valuable. The most important lessons learned by using CE in the present study are related to: The difficulties of creating a common design for an international river basin that will respect the methodological conditions (avoid attribute correlation), address the complexity of the socio-ecological systems and tackle the specificity of each country or river body. A common design that addresses different conditions will lead to making compromises like eliminating from the design important attributes which are crucial for explaining the systems complexity but which can create correlation or are difficult to explain to the general public in terms that are easily understood by everyone or setting attribute levels that will correspond to all the countries reference conditions. Regardless of the software used to produce the experimental design, which was SPSS for the Danube case-study, some changes in the choice sets will need to be done in order to avoid any irrational alternatives. The revision of the choice sets must be done with extreme care in order not to affect the orthogonality of the experiment. Expertise in the experimental design is essential for the whole study. The use of pictograms and colors seems to be very useful to explain the water quality levels to the general sample, but they must be tested for all age categories to see if all the visual elements are easily perceived and understood by the respondents (for example some visual elements can be more difficult to perceive by older population). The use of maps is helpful first for introducing the study area, to present scenarios and to create awareness among the respond over the fact that the subject of the evaluation (e.g. restoration project) refers to the area where they live, work, recreate etc. The level of maps must be carefully tested in order to send all the visual stimuli to the respondents providing the right level of detail and avoiding the overload of information. As the design is crucial for the credibility of the valuation exercise it is important to allocate enough time and resources to identify the right attributes and their level for the study through focus groups, interviews and pre-testing. Using face-to-face sampling had the following advantages and disadvantages: (+)Advantages: good representativeness of the sample for the decided criteria (urbanrural and gender ratios) but also age following the sampling routine; low refuse rates; easier to convince people to finish the questionnaire; operators had a strong inside of people’s motivations for choosing between alternatives Disadvantages: difficulties in handling all the additional materials (maps, water quality letter, payment cards) for only one operator; long time for applying the questionnaire (average 35 minutes); could include operator bias; time needed for operators debriefing meetings after each day of field work, a lot of time spent with traveling in different locations some of them difficult to access due to infrastructure problems. Finding reliable official data to undertake the socio-economical characterization of the local population sometimes seems an impossible task. Using older studies can sometimes help the process of identifying the representativity criteria, but in other situation giving up to some criteria seem to be the only solution in a study. 26 The time required to fill in a questionnaire in face-to -face surveys (an average of 35 minutes for about 35 questions) makes valuation expensive and time consuming. The time chosen for the survey implementation plays an important role. In the Braila Islands study the survey was implemented in November, which was ideally for the rural areas, when people are not so much involved in agricultural work but it was a problem for the operators due to the low temperatures as the surveys were applied on the street. The use of econometric software can raise problems without the proper training and expertise. As conjoint valuation are sometimes dependent on the respondents perception of the real conditions it is important to thoroughly inform people first about the current situation and to find suitable explanations for the attributes that need to be valued. Conjoint valuations will always raise questions: How do you know whether respondents would actually pay what they say they would? How can huge projects’ budgets be covered from household payments only? How can you make sure that the project results will mach the expectations of the population? How are the perverse effects of a policy/project considered in WTP valuations if these are only used to reflect improvements in the attributes considered for the project? These questions are difficult to address by carrying a one-shot evaluation, therefore complementary information has to support the studies results and has to reflect the changes in people’s attitudes and behavior that appear in time. People’s awareness is also dependent on the level of dissemination of studies result. The present study results will be disseminated through the reports sent to the National Water Authority which will include results in the River Basin Management Plan that will be subjected to public consultation. 27